A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 11:00 UTC
[Bugfix] Fix for the condition to accept empty encoder inputs for mllama (#17732)
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Gregory Shtrasberg committed
May 6, 2025 at 15:59 UTC
de906b95f9d0b9669da902785a9012ac96edd578
1 file changed
+1
-1
vllm/engine/llm_engine.py
+1
-1
@@ -2021,7 +2021,7 @@ class LLMEngine:
2021
if not prompt_ids:
2022
if prompt_type == "encoder" and model_config.is_multimodal_model:
2023
pass # Mllama may have empty encoder inputs for text-only data
2024
- if prompt_inputs["type"] == "embeds":
2024
+ elif prompt_inputs["type"] == "embeds":
2025
pass
2026
else:
2027
raise ValueError(f"The {prompt_type} prompt cannot be empty")